Why is my reaction time slower on a phone than on a computer?

These are arithmetic examples, not published phone-versus-desktop corrections. A stimulus may wait for a refresh, but its total presentation delay also involves rendering and the display itself. Input registration follows a separate pathway. Measurements across online platforms show that accuracy and precision vary across combinations of equipment and software. Anwyl-Irvine et al., 2020 Bridges et al., 2020

What does the displayed time include?

The recorded interval begins and ends at software events. It is intended to approximate the interval between seeing a cue and making a response, but the physical events can occur later than the software timestamps suggest. A laboratory study illustrates the size of this distinction: Woods and colleagues reported 231 ms measured simple reaction time and 213 ms after their equipment correction. Their 18 ms correction belongs to that apparatus, not to your phone. Woods et al., 2015

Touching a screen, clicking a mouse and pressing a key also require different movements. Changing the response changes the task even if the page looks identical. A choice-reaction experiment adds response selection and cannot be compared directly with a simple tap-on-green task. Woods et al., 2015

Why a smoother average is not necessarily a more accurate average

Accuracy means closeness to the true physical interval; precision means consistency. A device can be consistently delayed, producing a stable but offset score. Repeating the test helps estimate variability but does not remove a fixed hardware offset. That distinction is central to the experimental timing comparisons. Anwyl-Irvine et al., 2020 Bridges et al., 2020

What else changes when you change devices?

Screen size, posture, finger position and familiarity can change along with the hardware. Age differences reported in a controlled study cannot be used to infer that a hardware-induced score difference reflects biological ageing. Comparing yourself with another person's different device adds the same ambiguity. Woods et al., 2015 Woods et al., 2015

Reaction-time distributions also contain unusually slow responses. One lapse can move the mean substantially in a short test. Outlier rules and the choice of mean, median or fastest response therefore matter; selecting whichever summary looks best after the run undermines the comparison. Ratcliff, 1993

A useful way to compare your own runs

Keep a separate record for each device and input method. Compare like with like, including the number of attempts and the summary statistic. Repeating a fixed protocol is more interpretable than alternating devices and keeping the smallest score. Experimental work on trial counts shows that the appropriate amount of data depends on what is being estimated. Miller, 2023

Practice can also improve performance on repeated cognitive tasks. It is reasonable to expect familiarity to matter, but a general retest meta-analysis does not tell us the exact practice gain for this browser game. Nor does a large difference between two runs establish a reliable individual trait: the reliability of cognitive tasks must be measured rather than assumed. Scharfen et al., 2018 Hedge et al., 2017

There is consequently no evidence-backed rule such as subtracting the same delay from every phone result. Calibration requires measuring the actual presentation and input system. A refresh-rate specification alone is insufficient. Anwyl-Irvine et al., 2020 Bridges et al., 2020